Extract Figures from PDF Online
Automatically find embedded images and best-effort complete figure regions, or manually crop figures, charts, tables and diagrams. This free tool exports PNG, JPG or WebP while processing your PDF locally in your browser.
Choose a PDF
Practical guidance: start below 50 MB or 100 pages, then use a page range. Complex PDFs and memory-limited phones may need smaller ranges.
Advertisement
Automatic results
Detected complete figure regions
Best-effort candidates are built from rendered page structure and likely “Figure”, “Fig.” or “Table” captions. Choose Edit boundary to inspect, move or resize a candidate before adding it.
Embedded images
These are source raster objects, so a logo, mask or one panel may appear separately. Use a complete region when you need the composed figure.
Pages
Crop page 1
Draw with touch, pen or mouse. Drag inside the boundary to move it; drag a handle to resize. Zoom and pan for small figures.
No crop selected.
Review and download
Rename items, choose what to include, and download one file or a ZIP. Adding a crop does not trigger an immediate download.
No items added yet.
How to extract a PDF figure
- Choose your export format, quality and resolution, then open a PDF or the local sample.
- Use Auto-extract for source images and suggested complete regions, or Crop manually for exact control.
- Review each boundary. Move it, resize it, type numeric dimensions, include a caption if needed, and add it to the review queue.
- Rename and select the results you want, then download one image or a ZIP.
Embedded image vs figure crop: which should I use?
| PDF content | Embedded image | Complete figure crop | Recommended export |
|---|---|---|---|
| Raster photograph | Best when one source object contains the photo; keeps native pixels. | Use when labels, borders or a caption sit outside the photo. | JPG/WebP at 85–95%, or PNG for lossless editing. |
| Vector diagram or chart | Usually unavailable as one raster object. | Captures paths, labels and axes together from the rendered page. | PNG at 2× for web/slides; 4× for print. |
| Multi-panel or composite figure | May return separate panels, masks or backgrounds. | Captures the assembled A/B/C layout with shared labels. | PNG, typically 2×–4×. |
| Scanned page | May return the whole page scan. | Use manual crop to isolate the desired area. | JPG/WebP for photos; PNG when small text matters. |
| Caption required | Caption text is normally separate. | Enable likely caption or extend the boundary manually. | PNG for readable text. |
Resolution formula: output pixels = PDF points × DPI ÷ 72. Use 1× (72 DPI) for quick previews, 2× (144 DPI) for slides and normal web use, 3× (216 DPI) for high-density screens, and 4× (288 DPI) for print-oriented output. The source PDF still limits real detail.
Sample workflow: from PDF page to named figure
The one-click sample is created locally and contains a raster heatmap, a vector chart, and a captioned composite. It is safe to use as a quick browser test.
Before
Suggested boundary on page 2; resize handles remain editable.
After
Boundary: 420 × 260 PDF points
At 2×: 840 × 520 pixels
Filename: sample-vector-chart-p2.png
PDF figure extraction FAQ
Do my PDFs get uploaded or stored?
No. PDF document bytes are read and processed locally by PDF.js in your browser. The tool does not upload or store your document.
Can I extract one image instead of every image?
Yes. Set a page range, run automatic extraction, select only the results you want, and download the selected items. Manual mode can add a single crop to the same review queue.
What output formats are supported?
Crops can be exported as PNG, JPG or WebP. PNG is recommended for diagrams and text; JPG or WebP can be smaller for photographs. Embedded images keep their native pixel dimensions when that option is enabled, but PDF.js may still decode and re-encode the file.
Can the crop include a figure caption?
Yes. Enable Include likely caption before automatic detection, or resize a detected region or manual crop to include the caption. Caption matching is best effort and should be checked in the preview.
Does it work with scanned PDFs?
Yes, but a scan is usually one full-page raster image. Automatic mode may return the page image; use manual mode to isolate a figure. This tool does not perform OCR.
How are vector and multi-panel figures handled?
Vector charts and multi-panel layouts are not single embedded image objects. Best-effort region detection analyzes the rendered page, and manual mode lets you refine or draw a complete boundary before raster export.
Why do logos, masks or figure pieces appear separately?
PDFs can store one visible figure as several raster objects, masks, vector paths and text. The Embedded images list reflects those source objects; choose a detected complete region or manual crop to capture the assembled figure.
Can I process only selected pages?
Yes. Enter pages such as 1-3,5 or use Current page, then start automatic extraction. Cancel stops after the active page; Retry starts the chosen range again.
Which browsers are supported?
Current versions of Chrome, Edge, Firefox and Safari with JavaScript, Canvas, Pointer Events and Blob downloads are supported. Older browsers and restrictive in-app browsers may not work.
What are the file and page limits?
There is no fixed server limit because processing is local. As practical guidance, start below 50 MB or 100 pages, use a page range, and reduce scale on memory-limited phones. Complex PDFs may fail below those figures.
How does output resolution work?
At scale 1x, one PDF point becomes one output pixel, equivalent to 72 DPI. Output pixels equal PDF points multiplied by DPI divided by 72, so 2x is 144 DPI and 4x is 288 DPI.
Can I open password-protected PDFs?
Encrypted or password-protected PDFs are not supported. Remove the password locally, then try again.